March 14, 2009

Correlation and Causation

From Exurban Nation:
Rob_Dawg:
I used to think correlation was causation, then I took a statistics course and now I know better. Did the course help? I can never be sure.

My published articles are archived at iSteve.com -- Steve Sailer

54 comments:

Brett said...

http://xkcd.com/552/

rast said...

Boooo. That was a ripoff of a very very recent XKCD:

http://xkcd.com/552/

coldequation said...

That was probably lifted from xkcd:
http://xkcd.com/552/

If you hold your mouse over the cartoon, it says "Correlation doesn't imply causation, but it does waggle its eyebrows suggestively and gesture furtively while mouthing 'look over there'."

s_baghaii said...

I think the original source for this was the xkcd comic that is about a week older.

Guts Strongman said...

Same as:

http://xkcd.com/552/

PeterW said...

Or, http://xkcd.com/552/

Michael said...

Not sure who was there first, but I thought that originated with this webcomic:

http://xkcd.org/552/

YA said...

That's taken from ultra-popular geek webcomic XKCD (xkcd.com).

Muswell Hillbilly said...

That was taken from here:

http://xkcd.com/552/

Anonymous said...

A variation on this theme:

http://xkcd.com/552/

AMcGuinn said...

Originally from xkcd

Xenophon Hendrix said...

xkcd

Sean P. said...

Psst: http://xkcd.com/552/

Anonymous said...

There was probably a big uptick in conversations like this when this xkcd comic came out last week: http://xkcd.com/552/, but we'll never know for sure.

Anonymous said...

I'm pretty sure this is the origin of the line:

xkcd.com/552/

-Steve Johnson

Anonymous said...

For many people, learning that correlation is not causation might actually make them dumber since it will give them yet another weapon to reject ideas they find offensive.

What people should learn first (among other things) is that there is a difference between (1) how the world is; and (2) how we wish the world to be.

Anonymous said...

This seems like a ripoff of a recent xkcd comic. It's still funny, though.

Anonymous said...

http://xkcd.com/552/

AllanF said...

Done much better here:

http://xkcd.com/552/

Anonymous said...

did he steal that from XKCD or vice versa?

http://xkcd.com/552/

Anonymous said...

Somebody has been reading xkcd.

http://www.xkcd.com/552/

Mike said...

Here's where the joke originated:
http://xkcd.com/552/

Funny stuff.

ben g said...

stolen from xkcd.com?

http://xkcd.com/552/

Statsquatch said...

RA Fisher is not amused. If he hadn't died of cancer I am sure he would give xkcd an earful.

Anonymous said...

Wow, a lot of xkcd readers here!!!

Jim Bowery said...

Physics Without Causality - Theory and Evidence

Superdude said...

I refuse to believe this joke originally appeared in XKCD until we get twenty more comments telling us so.

Flamin' Nora said...

Is this all some big in-joke?

Steve Sailer said...

"Is this all some big in-joke?"

Not originally, but it's turning out to be one.

Colin Laney said...

Here's a link to the abstract of: Physics Without Causality - Theory and Evidence

Basically, some robust evidence on retrocausality in PK experiments and certain novel approches to quantum physics - namely the "transactional model" (brief summary and links here: http://bboyneko.livejournal.com/303972.html) - have created, not an acausal universe, but where causality is reversed in certain circumstances, flowing from future to present.

For those of you who are laughing out loud right now, may I invite you to go and review the results of the "double slit experiment", where the photons invariably "know" whether there is one slit or two awaiting them. Those are some pretty robust experimental results, and retrocausality (a final state interfering or influence the present moment) is certainly more plausible than some outgrowths of quantum math, especially the Many Worlds interpretation.

In passing, I note that Hume once brought the robustness of regular "causal" theory into doubt to great effect. So even if retroPK and the transactional interpretation of quantum theory are wrong in some way, there's still the possibility that they'll rouse some Kant from his dogmatic slumbers.

Also, if there is something to this view of things, then we are currently in a very strong version of the Anthropic Principle, where innumerable futures that haven't come into being yet are all competing to have the greatest effect on the present, so as to secure their own eventual existence. A sort of cosmic game of Survivor! Frederick Turner wrote a wonderful essay years ago called "Angels from the Time to Come", about just this struggle for existence. I do admit that it gives me a laugh to think that the work of the Ahthropic Principle is, under this dispensation, never done.

I do disagree with deeming this whole line of thought as 'Physics Without Causality", though. I prefer the term 'Retrocausality', since causality is conserved, it's just, um, postponed.

Anyway, funny joke from xkcd.

Anonymous said...

Correlates with http://xkcd.com/552

Curious about Krugman said...

Sorry to belabor the point, Steve, but is (or was) Krugman ever a member or part of the HBD Institute?

Sorry for asking again, but no one seems to have answered. I think you may be the only person who knows.

Thanks.

K said...

The funny part is that the remark "someone reads xkcd" was already used in the Exurban Nation blog post.

Randall Munroe is occasionally funny when he remembers to include his nerdy jokes rather than just liberal political sentiments.

Anonymous said...

I've never heard of the site; so what's with the xkcd.com divide?

Steve,
You have to do a poll of the people here to find out more about your readership.

DK said...

I get the joke, but I don't really have a rigorous definiton of cause. When I think about it, isn't cause something like 100% correlation of two events seperated by time, so isn't cause just a restricted type of very high correlation anyway? I'm not pretending to know. Others can help me get this straight. Its just that I hear people getting very excited about how correlation ins't cause as if they were in possesion of some great enlightenment. If the correlation between two types of events is very great then it doesn't matter if one causes the other or not. The fact is we know that one always implies the other *logically*, and and it doesn't really matter if one is strictly a cause of the other or not. For example if you know that walking down a certain dark alley automatically results in getting murdered, that's what is significant. Precise causes don't really matter. All that matters is strength of correlation.

OK if I'm wrong, someone please set me straight.

coldequation said...

Wow, there are so many xkcd readers here that the SPLC should probably put it on a list of hate comics.

Edward said...

It seems pretty obvious these comments were all sent in before the moderator published the first load.

Who else is really impressed that isteve readers have managed to come up with 22 completely original ways of telling us it was said first on xkcd?

Svigor said...

Maybe if a few more comments point to the xkcd link...

WHOOPS! Someone already pointed that out!

Remember, the appeal to authority is a classic logical fallacy. :)

Beastmaster said...

The original source for such a comment as that one was http://xkcd.com/552/.

Steve Sailer said...

There's got to be some even better meta-jokes here. Get to work!

Jim Bowery said...

For those not rolling on the floor laughing their asses off about Colin Laney's explanation, John Walker's musings may be of interest although it might also be of some interest that I could not get Mr. Walker interested in the theoretical work of the guys who formed the Boundary Institute. I'm not sure, but it seemed to me that 1996, about the time I approached Walker with Etter's "Link Theory", some guy named Sarfatti showed up in Walker's life and injected a lot of noise into the environment.

Anonymous said...

Steve,
Right after posting that I had never heard of the site, I showed my husband this post... he then proceeded to tell me it is "famous" and he spent the next 15 minutes sharing them with me and laughing at the "classics". BTW, his favorite was Shrodinger's Comic Strip or some such.

albertosaurus said...

There are really only two important ideas in elementary statistics: the Neymann-Pearson Lemma and the Central Limit Theorem. One for inferential statistics and one for descriptive statistics.

The "correlation isn't causation" statement is too elementary to require much discussion in a real stat class.

Even so it has escaped Al Gore that the correlation between CO2 and temperature could indicate causation in either direction.

Failure to understand the Central Limit Theorem has driven a lot of would be environmentalists to make absurd statements about "Hubbert's Peak". This doesn't mean that the oil won't eventually run out, just that Hubbert's Peak theory won't tell you when.

The most important statistics fallacy of all time is undoutedly the misuse of Principal Component Analysis in the MB98 paper that launched the whole Global Warming frenzy. PCM is a descriptive statistics technique that "smooths" time series data. Michael Mann - prmcipal author of MB98 - employed a novel variant of the accepted PCM procedure (see Wikipedia). The result of his innovation was the famous "hockey stick".

Mann admitted at the time that he was not a statistical expert and subsequently his new method has been rejected by virtually all real statisticians. Today the hockey stick is considered a mistake or a deliberate fraud. But the even after a decade of cooling the image of the hockey stick still haunts Washinton.

Obama wants to spend yet more billions on a problem based largely on discredited statistics.

The only example I can think of in using erroneous inferential statistics would be basing a conclusion on the efficiency of Italian versus Vatican mail delivery based on a sample size of one letter each. LOL.

Ronduck said...

Anonymous said...

For many people, learning that correlation is not causation might actually make them dumber since it will give them yet another weapon to reject ideas they find offensive.

What people should learn first (among other things) is that there is a difference between (1) how the world is; and (2) how we wish the world to be.


To continue your line of thought, I think education for most of the population should not extend beyond 8th grade. Most real "education" on makor issues such as economics, law, and history is formed by the popular press and not by schoolteachers. Also, it often takes ten years since the end of formal education for many people to shake of the liberal brainwashing they received, so the sooner the brainwashing ends the better. Hell, I would be a lot better off if my high school education ended at 16.

Once a person has learned to add, subtract, multiply, divide and read, it be better to send them home and honestly tell them that now they need to learn on their own from their daily experiences and personal efforts instead of waiting for someone to teach them.

Glaivester said...

If the correlation between two types of events is very great then it doesn't matter if one causes the other or not. The fact is we know that one always implies the other *logically*, and and it doesn't really matter if one is strictly a cause of the other or not. For example if you know that walking down a certain dark alley automatically results in getting murdered, that's what is significant. Precise causes don't really matter. All that matters is strength of correlation.

The problem is that correlation may be path-dependent. For example, let's say that there is a correlation between eating in restaurants and getting lung cancer. While the immediate conclusion might be that it is unsafe to eat in restaurants, it might turn out that the issue is simply that smokers tend to eat in restaurants more than non-smokers (and let's assume that they aren't smoking in the restaurant, it's just that people who are smokers prefer to eat out). That would then mean that there is nothing carcinogenic about going to a restaurant, so the correlation, while predictive of who might get lung cancer in a situation where all that you know is whether or not someone eats out, is not actually predictive if you are able to control for other variables, and is not predictive for the purposes of making decisions.

David Davenport said...

retrocausality (a final state interfering or influence the present moment) is certainly more plausible than some outgrowths of quantum math, especially the Many Worlds interpretation.
...


Why is retrocausality more plausible than Many Worlds?

Please explain.

Anonymous said...

Nerd: Fan of xkcd.com

Nerd amongst Nerds: Does not realize that the subject of this thread has now become "WHY on Earth are so many Sailerites xkcd.com readers when it seems to be kind of obscure and unrelated".

silver said...

There's got to be some even better meta-jokes here. Get to work!

Well, not meta, but as it pertains to "HBD"/race-realism: colonization is not causation.

David said...

Causality is like certainty. It requires 100% knowledge of all factors affecting all outcomes.

In short, it's a straw man. No one is omniscient.

Outside of mathematics, "cause" and "certain" and "knowledge" are probalistic.

The key is knowing when your knowledge is "certain enough" (i.e. probable enough) to act upon. Ask any surgeon.

"Corrolation vs. causality" really is "insufficient corrolation vs. sufficient corrolation" (i.e. sufficient for using as a basis for taking action in a given context).

If you wait to act until you're fully certain - or until you know A causes B in 100% of cases, past, present, and future - you will not be able fall in love, accept a job, befriend someone, eat in a restaurant, perform surgery, or even cross a busy street.

Anonymous said...

Ive just read this:

That's because you don't understand science. In science if you have one piece of tiny contrary evidence then your theory falls apart. The fact that you find one single British person cleverer then the Japanese person tells you that you cannot make a law out of your research and that your experiment is inconclusive. You have no idea how rigorous science is.

Posted by: Hulya Yadsan-Appleby


Pure comedy gold, you couldn't make it up!

It's from this thread. Ive tried posting a response but its not visible yet.

Ivy League Bastard said...

Correlation is correlated with causation.

Anonymous said...

This may or may not be the cause of this post:

xkcd.com/552/

Anonymous said...

Maybe his knowledge that correlation is not causation caused the stats class, is that what he's saying?

Rob Dawg said...

I had no idea I caused such a stir. When I made the quip on another blog I acknowledged the source as part of the conversation.